Regression Prediction of US Stock Prices Based on RBF Neural Network

The stock market is a dynamic nonlinear system filled with various types of noise, and accurately predicting it is a challenging task. This paper applies a Radial Basis Function (RBF) neural network for regression prediction of the stock price of Bank of New York Mellon in the US stock market. First, the study analyzes the time series characteristics of the stock and finds that during the study period, the Bank of New York Mellon stock price exhibited high volatility, with both the price fluctuation and amplitude revealing significant market instability. The dataset is then divided into training and testing sets, and normalization is applied to enhance the data's stability. An RBF model is constructed for prediction. The results are evaluated using metrics such as Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and the Coefficient of Determination <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(R^{2})$</tex>. Experimental results show that RBF outperforms traditional BP neural networks and genetic algorithm prediction models in terms of both prediction accuracy and computational efficiency. In particular, the RBF model shows significant advantages in RMSE, MAE, and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}$</tex>. RBF not only effectively addresses the slow convergence and local minima issues in the training process of BP neural networks but also demonstrates strong nonlinear mapping capabilities and fast learning convergence characteristics. Through validation with actual stock closing price data, the proposed model shows significant improvements in prediction accuracy and operational efficiency, offering an effective solution for stock price forecasting.